Herbaspirillum seropedicae signal transduction protein PII is structurally similar to the enteric GlnK Elaine Machado Benelli 1 , Martin Buck 2 , Igor Polikarpov 3 , Emanuel Maltempi de Souza 1 , Leonardo M. Cruz 1 and Fa ´ bio O. Pedrosa 1 1 Department of Biochemistry, Universidade Federal do Parana ´ , Curitiba, Brazil; 2 Department of Biological Science, Imperial College of Science, Technology & Medicine, Sir Alexander Fleming Building, Imperial College Road, London, UK; 3 Laborato ´ rio Nacional de Luz Sincrotron, Campinas, Brazil PII-like proteins are signal transduction proteins found in bacteria, archaea and eukaryotes. They mediate a variety of cellular responses. A second PII-like protein, called GlnK, has been found in several organisms. In the diazotroph Herbaspirillum seropedicae, PII protein is involved in sensing nitrogen levels and controlling nitrogen fixation genes. In this work, the crystal structure of the unliganded H. sero- pedicae PII was solved by X-ray diffraction. H. seropedicae PII has a Gly residue, Gly108 preceding Pro109 and the main-chain forms a bturn. The glycine at position 108 allows a bend in the C-terminal main-chain, thereby modifying the surface of the cleft between monomers and potentially changing function. The structure suggests that the C-terminal region of PII proteins may be involved in specificity of function, and nonenteric diazotrophs are found to have the C-terminal consensus XGXDAX(107–112). We are also proposing binding sites for ATP and 2-oxoglutarate based on the structural alignment of PII with PII-ATP/ GlnK-ATP, 5-carboxymethyl-2-hydroxymuconate iso- merase and 4-oxalocrotonate tautomerase bound to the inhibitor 2-oxo-3-pentynoate. Keywords: nitrogen regulation; PII X-ray structure; crystal packing, Herbaspirillum seropedicae;GlnK. Control of nitrogen metabolism in many bacteria utilizes a conserved mechanism of intracellular signalling to regulate patterns of gene expression and enzyme activity necessary for adapting to changes in the quality and abundance of nitrogen sources. The NifA protein is the transcriptional activator of nitrogen fixation (nif) genes in the majority of diazotrophs within the Proteobacteria. In several of these organisms, nifA expression is controlled by the general nitrogen regulation Ntr system, which, in turn, is controlled by the state of the glnB product, the PII protein. Under nitrogen excess, PII interacts with NtrB resulting in the dephosphorylation of the transcriptional activator NtrC-P and diminished nifA expression. Under limiting nitrogen, PII is uridylylated by GlnD and this allows NtrB to phosphorylate NtrC. In the c-subdivision of the Proteo- bacteria, nif gene expression is regulated by NifA and NifL: under high ammonium or oxygen levels NifL inhibits NifA activity, whereas under nitrogen limiting conditions and low oxygen NifA is active. In K. pneumoniae GlnK, a paralogue of PII, interacts with the NifL–NifA complex, to relieve NifA inhibition by NifL [12,13,16,]. In Azotobacter vinela- ndii only the GlnK protein is present and it controls the activity of NifA by the interaction with NifL and the complex NifL–NifA is sensitive to 2-oxoglutarate levels [20]. Although extensively studied in bacteria, PII-like proteins are present in all three kingdoms of life. For recent reviews see Ninfa & Atkinson [24], Thomas et al. [33] and Mag- asanik [21]. In Herbaspirillum seropedicae, a member of the bsubdi- vision of Proteobacteria, the glnAntrBC and glnB genes have been identified [6,26], suggesting that an Ntr PII-dependent signal transducer cascade senses the nitrogen levels in this organism. In H. seropedicae, nifA expression is also dependent on phosphorylated NtrC (NtrC-P), but NifL has not been found. However, the activity of NifA is known to be controlled by the PII protein, as in Azospirillum brasilense, a member of the a subdivision of the Proteobac- teria [2,3]. The mechanism involved in this control is not known. Souza et al. [30] observed that the activity of a H. seropedicae N-terminal domain-truncated NifA (DNTD) was independent of ammonium levels, suggesting that the N-terminal domain (NTD) plays a role in the control of NifA activity by ammonium. Arsene et al.[3] made a similar observation in A. brasilense and suggested that PII-UMP may interact with the NTD of NifA to change its activity. The residue Tyr18 from the NTD of NifA seems to be involved in the interaction between PII and NifA [2]. PII proteins interact directly with a variety of ligands, including ATP and 2-oxoglutarate. The structure of the EcPII protein and the paralogue EcGlnK have been solved in the presence and absence of ATP [7,9,35,36]. Here we report the crystal structure of unliganded H. seropedicae PII (HsPII) at 2.1 A ˚ resolution and compare this with the available structures from E. coli. Although in amino-acid Correspondence to E. Machado Benelli, Department of Biochemistry, Universidade Federal do Parana ´ , C. Postal 19046, Curitiba, Brazil. E-mail: benelli@bio.ufpr.br Abbreviations: NtrB, nitrogen regulation protein B; NtrC, nitrogen regulation protein C; GlnD, uridylylating enzyme; NifA, nitrogen fixation protein A; NifL, nitrogen fixation protein L; EcPII, Escheri- chia coli glnB product; EcGlnK, Escherichia coli glnK product; HsPII, Herbispirillum seropedicae glnB product; KpPII, K. pneumoniae glnB product; KpGlnK, K. pneumoniae glnK product. (Received 29 January 2002, revised 15 February 2002, accepted 22 May 2002) Eur. J. Biochem. 269, 3296–3303 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03011.x sequence HsPII shows higher identity to EcPII than EcGlnK, distinct structural differences are evident, placing HsPII closer to the unliganded and ATPbound forms of EcGlnK in three-dimensional structure. We suggest a correlation of the structural differences with the specialized functions of PII-like proteins in diazotrophs. It seems that function may be related to conformational flexibility exhibited by PII and GlnK proteins, as indicated by a comparison of crystal packing arrangements seen in several different crystal forms of PII-like proteins [7,9,35]. Changes in EcPII structure associated with ATP binding support this view and indicate that C-terminal structures can be ligand dependent [35]. When EcPII is bound to ATP the C-terminal structure is similar to that in unliganded EcGlnK [36] and unliganded HsPII (this paper). We note similarities in quaternary and subunit tertiary structure with other proteins, unrelated to PII by amino-acid sequence, that interact with a-ketoacids, suggesting the existence of a family of a-ketoacid interacting proteins. EXPERIMENTAL PROCEDURES Protein purification HsPII protein was overexpressed in E. coli RB9065kDE3, a glnB glnD mutant background lysogenized with kDE3 for T7 RNA polymerase production and purified as described by Benelli et al. [5]. The purified HsPII protein was dialysed in a buffer containing 10 m M Tris/HCl pH 8.0, 50 m M NaCl, 20% glycerol and 0.1 m M EDTA and concentrated in a Centricon-3 filter prior to crystallization. Crystallization Crystallizations used either the sitting or hanging drop vapour diffusion method at 18 °C in Limbro tissue culture plates. An initial Hampton crystallization screen of both native and N-terminal hexa histidine-tagged HsPII yielded promising microcrystals. Conditions were optimized by addition of a number of additives [10]. HsPII protein (14 mgÆmL )1 )andHis 6 –PII protein (13 mgÆmL )1 )inTris/ HCl 10 m M pH 8.0, NaCl 50 m M , glycerol 20% and EDTA 0.1 m M were used in crystallization experiments. A tetragonal crystal form of native PII was grown from hanging drops containing protein solution mixed in a 1 : 1 ratio with well solution (15.8% ethyleneglycol). A trigonal crystal form was grown by vapour diffusion in sitting drops. The reservoir solution contained 0.1 M sodium acetate pH 4.6, 30% methylpentadiol and 0.15 mgÆmL )1 of dextran sulfate. The drops contained 1 lLofprotein solution and 1 lL of reservoir solution. The orthorhombic crystal form grew, using the hanging drop method, in 30% methylpentadiol, 0.1 m M sodium cacodylate pH 6.5 and 0.2 m M magnesium acetate. Initial tests on a copper rotating anode revealed diffraction to 3 A ˚ from the tetragonal and trigonal crystal forms (Table 1). Crystals of His 6 –PII were obtained by the hanging drop method at 18 °C. The reservoir solution contained 0.5 mL of 0.1 M sodium citrate pH 6 and 10% PEG 6K and the drop contained 1 lL of protein solution (13 mgÆmL )1 )and1lL of reservoir solution. The His 6 –PII crystal form diffracted to 6 A ˚ with the rotating anode source, and was not further characterized. Table 1. Summary of X-ray data collection and crystallographic refinement statistics. Data collection a,b Space group P2 1 2 1 2 1 P3 2 21 P4 3 2 1 2 Unit cell dimensions a ¼ 78.41 A ˚ ,b¼ 82.36 A ˚ ,a¼ b ¼ 121.74 A ˚ ,c¼ 65.24 A ˚ a ¼ b ¼ 88.81 A ˚ ,c¼ 116.91 A ˚ c ¼ 100.95 A ˚ a ¼ b ¼ 90°, c ¼ 120° Solvent content (%) 68 68 61 Max. resolution (A ˚ ) 2.1 3.0 3.2 Unique reflections 36523 21170 8163 Redundancy 3 3 4 Completeness (%) 94 (95) 98 (99) 100 (100) Average I/rI 13 (1) 12 (2) 14 (4) R ¼ S|I ) <I>|/S|I| 0.057 0.078 0.187 Refinement in orthorhombic crystal form c,d,e Data range (A ˚ ) 13.0–2.1 Reflections (F > 0) 36331 Completeness (%) 94.4 Reflections in free set 1820 Non-H atoms 4313 Residues 560 Rms bond lengths (A ˚ ) 0.018 Rms bond angles (deg) 0.044 Rms B-factors for bonded atoms (A ˚ 2 ) 4.2 R free (%) 27.2 R cryst (%) 20.3 a Values in parentheses correspond to the highest resolution shell; 2.15–2.10 A ˚ (2415 reflections) for the orthorhombic form; 3.05–3.00 A ˚ (815 reflections) for the trigonal form; 3.25–3.20 A ˚ (393 reflections) for the tetragonal form. b The resolution cut-off was defined so that 50% of reflections in the highest resolution shell had I > 3 r. c Rms deviations in bond lengths and angles are given from ideal values. d R cryst ¼ S||F o |–|F c ||/S|F o |. e R free is as for R cryst but calculated for a test set comprising 1820 reflections not used in the refinement. Ó FEBS 2002 Structural similarities between PII and GlnK (Eur. J. Biochem. 269) 3297 Data collection and processing A summary of the data collection and refinement statistics is given in Table 1. Diffraction data were collected from a single crystal of each form at 120 K using a 30-cm MAR imaging plate detector system on a RIGAKU RU-200B generator with a copper anode and double focusing mirrors. A2.1-A ˚ data set on the orthorhombic crystal form was collected at 120 K using synchrotron radiation at a wavelength of 1.38 A ˚ , using a MAR 345 imaging plate on the protein crystallography beamline [28,29] at the Brazilian National Synchrotron Laboratory (Campinas, Brazil). The crystal initially diffracted to 1.9 A ˚ , but the high resolution reflections gradually decayed during data collec- tion. The diffraction data were consistent with space group P2 1 2 1 2 1 , with the cell parameters a ¼ 78.41 A ˚ , b ¼ 82.36 A ˚ , c ¼ 100.95 A ˚ . The data were integrated, reduced and scaled using DENZO and SCALEPACK [25], respectively. Intensities were then converted to structure factors using the method of French & Wilson [11] as implemented in TRUNCATE [8]. Structure solution and refinement The structure of HsPII was solved in three space groups by molecular replacement in AMORE [23]. Selected crystallo- graphic data are given in Table 1. The complete EcPII monomer structure (PDB accession no. 2PII; [7]) and a truncated model lacking the uridylylation site loop and the C-terminal tail, residues 40–54 and 96–112, respectively, were both used as search models to solve the trigonal crystal form. Both monomer and trimer forms, generated by the crystallographic threefold axis in space group P6 3 ,wereused as search models. All calculations performed used 10 to 4 A ˚ data. Only when the trimer was used as a search model did the first peak in the cross rotation function correspond to the correct solution. A solution could not be found with the entire monomer structure, only with the truncated mono- mer model. Initial refinement of the whole model included noncrystallographic symmetry averaging and yielded a crystallography R-factor of 37%, the electron density map calculated at this stage indicated that residues 38–51 and 104–112 were not in correct positions. Model building was subsequently carried out on the truncated model only. The electron density for the rest of the protein was well defined; therefore it was possible to substitute all EcPII residues with the corresponding HsPII residues. Electron densities for residues 38 and 39 were so poor that they both had to be removed. Additional electron densities were apparent for two residues preceding Asp54 and five after Val96. The current model including residues 1–37 and 52–110 was obtained after a few rounds of model adjustment followed by refinement in REFMAC [22]. The tetragonal crystal form was solved using the trigonal HsPII model after the first build in which all the amino acids different from EcPII were changed. This model included residues 1–37 and 54–96. The structure of the orthorhombic crystal form was solved using the trigonal HsPII containing residues 1–35 and 55–107. Molecular replacement, including rotation and translation functions followed by rigid body refinement, was carried out using 10 to 3.3 A ˚ data and resulted in an R-value of 39.6% and correlation coefficient of 60.3%. Refinement was carried out using the program REFMAC [22] from the CCP 4 suite of the program [8]. Eighty cycles of positional and B-factor refinement of the molecular replacement model against all the data between 10 A ˚ and 2.1 A ˚ resolution resulted in a model with R cryst 30.0% and R free 36.1%. Model building was carried out using the programe O [18]. The orthorhombic HsPII model was built into 2F o ) F c and F o ) F c difference maps, residues were placed in well defined 2 r electron density maps. Eleven cycles of model building and refinement resulted in an R- factor of 23.1%. and R free of 29.8% In the last cycle, 125 molecules of water were added and the R-factors dropped to 20.3 and 27.3%, respectively. The final model comprises residues 1–37 and 51–112 (monomer A), 1–36 and 43–107 (monomer B), 1–36 and 57–112 (monomer C), 1–37 and 50–112 (monomer D), 1–35 and 57–105 (monomer E) and 1–35 and 57–112 (monomer F). The residue Lys68 is placed as Ala in chains B, D and F because the electron density of the lateral chain of Lys was not observed in these chains. The stereochemical quality of the final model of the HsPII protein was verified by PROCHECK [19]. The coordinates were deposited in the Protein Data Bank as the code 1HWU. RESULTS AND DISCUSSION Overall structure HsPII was overproduced and purified from E. coli and found to be a trimer of 36 kDa in solution, as are the EcPII and EcGlnK proteins [24]. The crystal structure was solved by molecular replacement using EcPII as the search model (see Materials and methods). Several different crystal forms of HsPII were grown (Table 1). The structural model was obtained from the orthorhombic crystals which diffracted at 2.1 A ˚ . Monomers of the HsPII trimer are accommodated around a central threefold axis (Fig. 1A). The core of the HsPII monomer has a double bab motif (Figs 1B and 2A). The structural scaffold (the b strands, the a helices and the short B-loop) is well conserved in available PII-like struc- tures (Fig. 1B). Major differences amongst structures are in the T-loop (which contains the uridylylation site, Tyr51) and C-loop. HsPII is similar to EcGlnK, EcGlnK-ATP and EcPII-ATP in its C-loop (Figs 1B and 2A). The b strands of the bab motif line the central cavity of the HsPII trimer, with the a helices at the periphery of the molecule (Fig. 1A). The bottom edge of the central cavity is negatively charged (Fig. 2B, part i) owing to the presence of Glu97 (Ala in EcPII and Gln in EcGlnK) and Glu95. The entrance of the central cavity is partially restricted by Gln94 with lateral chains directed towards its interior. Gln94 is substituted by Phe in EcPII and Ala in EcGlnK. The entrance from the top is restricted by Thr31, whose lateral chain is oriented to the interior of the cavity. The interior wall of the cavity is largely hydrophilic. Most of the intersubunit interactions that maintain the EcPII and EcGlnK trimers occur between conserved residues and are therefore also preserved in the HsPII structure, for example between Lys34 and Glu32 and Lys60 and Glu62 or Asp62 (in EcGlnK) (Fig. 2A,B, part ii). The salt bridge between residues Lys2 and Glu95 in EcPII appears to be substituted by Lys2 and Asp97 in HsPII. Furthermore, the interaction between residues Asp71 and Arg98 of different chains seen 3298 E. Machado Benelli et al. (Eur. J. Biochem. 269) Ó FEBS 2002 in EcPII does not exist in the HsPII structure. These residues are substituted by Glu and Gln, respectively. The lateral cleft created in the interface of each monomer of HsPII (Figs 1A and 2B, part iii) is similar to that observed in EcGlnK but smaller than in EcPII. In HsPII the clefts are partially obstructed by C-terminal sequences. The bend of the main chain at Gly108 pushes residues Pro109, Asp110, Ala111 and Val112 into the cleft (Fig. 2B, part ii). Around the lateral cleft in the HsPII protein there is a salt bridge between Asp66 and Lys68 which does not appear in EcPII or EcGlnK (Fig. 2B, parts i and iii). This bridge is close to the C-terminal region and might mediate the interactions between PII and its receptors. Single amino- acid modifications of EcPII protein around the cleft produced mutant proteins (residues Thr83, Gly89 and Lys90) with impaired binding of the ligands 2-oxoglutarate and ATP [17] (Fig. 4C). None of the C-terminal residues seem to interact directly with ATP in the EcGlnK or EcPII proteins, for which structures of the complexes with ATP are available [35,36]. However these C-terminal residues are closer to the lateral cleft in EcGlnK compared to EcPII and might therefore influence binding of ATP indirectly. It is reasonable to propose that the structure of the C-terminal region is important for effector binding to PII, althought the effector need not directly interact with the C-terminal sequence (discussed below). The HsPII protein requires 2-oxoglutarate for uridylylation by the GlnD protein whereas the EcPII requires both 2-oxoglutarate and ATP, and the affinity constant for 2-oxoglutarate binding to HsPII is considerably higher than that of EcPII [5]. The bottom face of the HsPII trimer comprises mainly negatively charged residues (Fig. 2B, part i). Positive charges are located around the B-loop, which is probably involved in ATP interactions. In this region, HsPII Arg101 and Arg103 are separated by the lateral chain of Ile102 whereas in EcPII these residues are closer. This may explain why in the presence of excess 2-oxoglutarate K act for ATP binding to HsPII is higher (100 l M ) than that for EcPII ( 3 l M ) [5,17]. The T- and C-loops The T-loop of PII-like proteins frequently includes a tyrosine which is the site of uridylylation. Where structures of the T-loop are available for EcPII and EcGlnK, crystal packing contacts appear to stabilize the T-loop in an artificially ordered conformation. In HsPII the part of T-loop that could be built shows a high temperature factor, and is exposed to the solvent. In the orthorhombic HsPII crystal there are two PII trimers per asymmetric unit. As Fig. 1. Ribbon diagrams of the trimeric HsPII (A) and monomeric HsPII, EcPII, EcPII- ATP, EcGlnK and EcGlnK-ATP (B). (A) A ribbon diagram of the structure of the trimeric HsPII, each chain in a different colour. The b sheets of the bab motif line the central cavity of the trimer with the a helices at the periph- ery. (B) Ribbon diagrams of the monomers of HsPII (i), EcPII (ii), EcPII-ATP (iii), EcGlnK (iv) and EcGlnK-ATP (v). Secondary struc- tures are colour coded: green b sheets, b1–4, blue helices, a1–2 and 3 10 helix and orange loops. The monomers share the same bab motif with the major structural differences residing in the loops T and C. Ó FEBS 2002 Structural similarities between PII and GlnK (Eur. J. Biochem. 269) 3299 packing contacts are different for each monomer in these two trimers, the monomers were refined independently. The final rmsd values for the overlay of all atoms of the two trimers was 0.42 A ˚ . Electron density in all HsPII monomers to residues 38–53, which are within the T-loop, were not completely visible and the C-loop could be built in four of the six monomers (monomer A, C, D and F) present in unit of cell (see Experimental procedures). Those residues of the HsPII T-loop that can be traced represent a conformation unaffected by crystal packing contacts and are presumably in the preferred conformation of the T-loop as exists in the absence of interacting ligands such as ATP and 2-oxoglut- arate. The limited amount of HsPII T-loop that is structured shows significant conformational differences compared to those sequences ordered by packing contacts in the crystals of EcPII and EcGlnK (Fig. 1B). This implies that changes in conformation across much of the T-loop are possible during the normal functioning of the PII-like proteins [1,35]. The K. pneumoniae glnK product (KpGlnK) and EcG- lnK proteins function to relieve NifL inhibition of NifA activity under nitrogen-limiting growth conditions. Arcon- deguy et al. [1], investigated the importance of the KpGlnK T-loop residues 43, 52 and 54 on the control of K. pneu- moniae NifA activity. Both EcGlnK and KpGlnK proteins have high sequence identity to EcPII. However, EcPII expressed from the chromosome is unable to substitute for the GlnKs with respect to NifLA [13,16]. Arcondeguy et al. (2000) suggested that residue 54 is the most important residue in the T-loop for distinguishing between PII and GlnK in controlling NifL activity. Residue 54 in HsPII is aspartate, as in K. pneumoniae glnB product (KpPII) and EcPII. However HsPII differs functionally from EcPII and KpPII, and is able to activate NifA in an E. coli background containing NifL when expressed from a low copy number plasmid, as does EcGlnK, but not EcPII or KpPII (A. C. Bonatto, E. M. Souza, F. O. Pedrosa & E. M. Benelli, unpublished results). This suggests that some determinants of functionality that distinguish PII from GlnK must reside outside the T-loop. Consistent with this a second HsPII-like protein has been discovered, with the same T-loop sequence as the HsPII studied here (L. Noindorf, M. B. Steffens, E. M. Souza, F. O. Pedrosa & L. Chubatsu, unpublished data). This protein was called GlnK because it has higher identity to EcGlnK than EcPII and it is encoded by a glnK gene which is located on the glnKamtB operon. The HsPII and HsGlnK proteins are 78.6% identical and 93.75% similar and one of the seven different amino acids is in the C-terminal (Pro109 HsPII is substituted by Lys109 HsGlnK). Despite, the high homology between these proteins they are functionally different. The H. seropedicae glnB mutant has normal GS activity and biosynthesis but it is unable to fix nitrogen, suggesting that in vivo HsGlnK is unable to substitute HsPII [6]. The C-terminal structure of PII The structure of the C-terminal region of HsPII could be entirely built only for one of the monomers in the asymmetric unit. In contrast to the C-terminal region of EcPII, which contains a bsheet, the C-terminal of HsPII contains a turn of Fig. 2. Alignments of the HsPII with EcPII and HsPII with EcGlnK amino-acid sequences (A) and molecular surface of the HsPII trimer (B). (A)AlignmentsoftheHsPIIwithEcPII and HsPII with EcGlnK amino-acid sequences. The identity (73%) and similarity (86%) of HsPII to EcPII is higher than HsPII to EcGlnK (67% and 76%, respectively). Secondary structural elements are labelled above and below the sequence. The a helix, b strands, 3 10 helix and loops are coloured in green, blue, dark green and black, respectively. (B) Molecular surface of the HsPII trimer colour-coded with acidic residue side-chains in red, basic side-chains in blue and others in white. T-loop residues 37–51 are not included. Residuesreferredinthetextarelabelledonthe monomer (i) The negatively charge bottom face of the trimer; (ii) the top face of the trimer and (iii) molecular surface of the lateral cleft of the HsPII trimer. The salt bridge between Asp66A and Arg68A, located close to the C-terminal region, may mediate interaction betweenPIIanditsreceptors. 3300 E. Machado Benelli et al. (Eur. J. Biochem. 269) Ó FEBS 2002 a3 10 helix as does EcGlnK, EcGlnk-ATP and PII-ATP (Fig. 1B). Although the identity (73%) and similarity (83%) of HsPII to EcPII is higher than that of HsPII to EcGlnK (67% and 76%, respectively), HsPII is structurally closer to EcGlnK or EcGlnK-ATP and EcPII-ATP (Figs 1B and 2A). The amino-acid sequence from residues 106–112, encoding part of the C-loop of PII-like proteins is only partly conserved (Fig. 3A). EcPII has a sequence of four negatively charged amino acids in this region (residue 106– 109), EcGlnK three residues (106, 108 and 109), whereas HsPII contains only two negatively charged amino acids at positions 106 and 110. The rmsd values in Ca positions obtained from superposition of the core (residues 1–35 and 56–95) were 0.58 A ˚ for HsPII-EcPII and 0.55 A ˚ for HsPII- EcGlnK. The rmsd values in Ca obtained for the superpo- sition of the C-terminal segments (residues 95–112) were 0.91 A ˚ and 0.43 A ˚ for EcPII and EcGlnK, respectively, establishing that HsPII is structurally closer in this region to the EcGlnK protein (Fig. 1B). The structural relatedness of HsPII and EcGlnK in their C-terminal regions may explain why HsPII is functionally similar to the KpGlnK and EcGlnK proteins. HsPII and KpGlnK are involved in the control of the NifA activity, as discussed above, whereas EcPII and KpPII are not [6,12,1,16,20]. Recent structure determination of the EcPII protein with ATP bound has shown that its C-terminal sequences can adopt a conformation very close to that of the unliganded EcGlnK [35,36] (Fig. 1B). The C-terminal part of unligan- ded HsPII, preferentially adopts the structure seen in unliganded EcGlnK (Fig. 1B). Although EcPII can adopt two different structures in its C-terminus depending upon its ligation state, we have no evidence for this in HsPII. Nevertheless ligand induced structural changes may well influence the functioning of the C-terminus of HsPII. Amino-acid sequence alignment of a C-terminal region (residues 106–112) (Fig. 3A) amongst PII proteins indicates that this region is distinctly more conserved amongst nonenteric diazotrophs (50% identity as opposed to 16% between E. coli and H. seropedicae) suggesting similarity of function. As with HsPII, the A. brasilense PII protein also activates NifA [2,3]. Residues Gly108, Asp110 and Ala111 are present in the PII proteins of H. seropedicae, Rhodo- bacter capsulatus, Rhodospirillum rubrum, Rhodobacter sphaeroides, Bradyrhizhobium japonicum, A. brasilense, Rhizobium leguminosarium and Azorhizobium caulinodans. The residue Thr107 is present in the majority of these organisms. On the other hand, the PII C-terminal sequences are highly conserved between E. coli and K. pneumoniae. These observations indicate that these proteins can be divided into two classes. The enteric organisms share the C-terminal sequence EDDAAI. In nonenteric diazotrophs the C-terminal consensus is XGXDAX (Fig. 3A). It seems the glycine at position 108 of the latter class allows a bend in the C-terminal main-chain, thereby modifying the surface of the intermonomer cleft and changing functionality. The contribution of these residues to HsPII function is under investigation. Relationship of PII to GlnK Jack et al. [16] aligned PII and parologue proteins from several organisms and found five residues (positions 3, 5, 52, 54 and 64), which distinguish GlnK from PII proteins. PII proteins contain Lys3, Glu5 or Asp5, Met52 or Val52, Asp54 and Val64. In contrast, in GlnK proteins these amino acids are: Leu3 or Ile3, Thr5, Met5 or Ile5, Ser52 or Ala52, Ser54 or Asn54 and Ala64. In HsPII three of these residues are identical to those of PII proteins: Met52, Asp54 and Val64; one (Thr5) is found in GlnK proteins (Fig. 2A). However, these alignments included only PII and paralogue proteins of the organisms from the a) and c-subdivision of the Proteobacteria. We have constructed a phylogenetic tree of PII and paralogue proteins including HsPII and the PII and paralogue proteins from Azoarcus, another member of the b-subdivision of Proteobacteria using CLUSTALX [34] (Fig. 3B). In this tree there are two groups of proteins: PII and GlnK, separated according the Proteobacteria subdi- visions. However, the HsPII was not included in either group, which emphasizes the special nature of HsPII and is consistent with the particular structural relationship HsPII bears to EcPII and EcGlnK. Structural alignment of H. seropedicae PII protein to others proteins Structural alignment of HsPII protein using the DALI program [14,15] showed that it has a relatedness to several Fig. 3. Alignment of the C-loop residues of PII proteins from nonenteric (the first nine sequences) and enteric bacteria PII and GlnK proteins (the remaining four sequences) (A) and phylogenetic tree of the PII and paralogue proteins in the proteobacteria (B). (A) The residues in blue show the conserved residues of the XGXDA motif in the nonenteric and enteric bacteria. Abbreviations are: Hs, H. seropedicae;Ab,Azo- spirillum brasilense;Ac,Azorhozobium caulinodans;Bj,Bradyrhizobium japonicum;Rl,Rhizobium leguminosarum;Rm,Rhizobium meliloti;Rr, Rhodospirillum rubrum;Rs,Rhodobacter sphaeroides;Rc,Rhodobacter capsulatus;Kp,K. pneumoniae;Ec,E. coli. (B) Phylogenetic tree of the PII and paralogue proteins in the proteobacteria. This shows two major groups with HsPII outlying these. The tree was calculated by the CLUSTALX program [34]. Ó FEBS 2002 Structural similarities between PII and GlnK (Eur. J. Biochem. 269) 3301 other proteins that possess a double bab fold, as nucleotide diphosphate kinase, RNA binding protein, ribosomal protein, allosteric domain of the regulatory subunit of aspartate transcarbamylase, a viral transcriptional regulator and procarboxypeptidase B [9]. Additionally, we found HsPII aligned with the enzymes: 5-carboxymethyl-2- hydroxymuconate isomerase and 4-oxalocrotonate tautom- erase. These proteins are involved in the isomerization of a-keto acids [31,32]. The superposition of HsPII with 4-oxalocrotonate tautomerase protein bound to 2-oxo- 3-pentynoate, an inhibitor of a-keto acid isomerization, suggests that the 2-oxoglutarate may bind around the lateral cleft region of PII (Lys90 and Arg101, from different monomers). Comparisions between HsPII, EcGlnK-ATP or EcPII-ATP and 4-oxalocrotonate tautomerase bound to the inhibitor 2-oxo-3-pentynoate [32] suggests that although the ATP and 2-oxoglutarate binding sites in HsPII are in the lateral cleft, they are not superimposed (Fig. 4). The suggested position of the 2-oxoglutarate binding-site is consistent with biochemical data that show that mutations in residues: G37, R38, Q39, K40, T83, G84, G89 and K90 affected the 2-oxoglutarate binding to EcPII [17]. Although, Xu et al. [36], suggested that 2-oxoglutarate could bind to the T-loop to stabilize this flexible loop, the present model shows that it is possible that 2-oxoglutarate can bind in the lateral cleft close to two Arg residues as in 5-carboxymethyl-2-hydroxymuconate isomerase and 4-oxalocrotonate tautomerase. In the isomerases the binding site also contains a proline residue involved in the catalysis; this proline is not present in HsPII. It is known that the affinity of E. coli PII for either ATP or 2-oxoglutarate is dependent on the other ligand, implying that each ligand causes a conformational change to increase acceptance of the second ligand [17]. ACKNOWLEDGEMENTS This work in part was carried out at the Departments of Biology and Biophysics, ICSTM and with Anne Harper, Madeleine H. Moore and Johan P. Turkenburg in the Protein Structure Group, University of York. We thank Silvia Onesti, Xiaodong Zhang and Marshall G. Yates for their constructive suggestions and David Ollis for the coordinates of the E. coli PII protein bound to ATP. This work was supported by CNPg, PRONEX/MEC and BBRSC. Fig. 4. Model to ATP and 2-oxoglutarate binding sites in HsPII protein. (A) Diagram of a Ca trace overlay of HsPII (orange) with CHMI (cyan) [31]. A top view of the trimers similar in orientation to Figs 1A and 2Bii. The different views of the proposed 2-oxoglutarate and ATP-binding sites in HsPII protein are shown in (B), (C) and (D). (B) Position of ATP and 2-oxo-3-pentynoate in the lateral cleft of HsPII. (C) ATP molecule and the neighbouring amino-acid residues. (D) 2-oxo-3-pentynoate molecule and the neighbour amino-acid residues. Location of ATP and 2-oxo-3- pentynoate was modelled using HsPII, EcGlnK-ATP, EcPII-ATP and 4-oxalocrotonate tautomerase-2-oxo-3-pentynoate structures using DALI and LSQKAB [8,14,15]. 3302 E. Machado Benelli et al. (Eur. J. Biochem. 269) Ó FEBS 2002 REFERENCES 1. Arcondeguy, T., Lawson, D. & Merrick, M. (2000) Two residues in the T-loop of GlnK determine NifL-dependent nitrogen control of nif gene expression. J. Biol. Chem. 275, 38452–38456. 2. Arsene, F., Kaminski, P.A. & Elmerich, C. (1999) Control of Azospirillum brasilense NifA activity by P (II): effect of replacing Tyr residues of the NifA N-terminal domain on NifA activity. FEMS Microbiol. Lett. 179, 339–343. 3. Arsene, F., Kaminski, P.A. & Elmerich, C. (1996) Modulation of NifA activity by PII in Azospirillum brasilense: evidence for a regulatory role of the NifA N-terminal domain. J. Bacteriol. 178, 4830–4838. 4. Benelli, E.M., Buck, M., Bonatto, A.C., Souza, E.M., Moore, M., Harper, A., Yates, M.G. & Pedrosa, F.O. (2000) Role of PII as a signal of nitrogen levels in Herbaspirillum seropedicae PII protein relieves NifA inhibition by NifL in a enteric background. Fourth European Nitrogen Fixation Conference, pp. 247. Seville, Spain. 5. Benelli,E.M.,Buck,M.,Souza,E.M.,Yates,M.G.&Pedrosa, F.O. (2001) Uridylylation of the P II protein from Herbaspirillum seropedicae. Can J. Microbiol. 47, 309–314. 6. Benelli, E.M., Souza, E.M., Funayama, S., Rigo, L.U. & Pedrosa, F.O. (1997) Evidence for two possible glnB-type genes in Her- baspirillum seropedicae. J. Bacteriol. 179, 4623–4626. 7. Carr, P.D., Cheah, E., Suffolk, P.M., Vasudevan, S.G., Dixon, N.E. & Ollis, D.L. (1996) X-Ray structure of the signal trans- duction protein PII from Escherichia coli at 1.9 A ˚ . Acta Crystal- logr. D52, 93–104. 8. CCP4 (1994) CCP4 suite: programs for protein crystallography Collaborative Computational Project, Number 4. Acta Crystal- logr. D50, 760–763. 9. Cheah,E.,Carr,P.D.,Suffolk,P.M.,Vasudevan,S.G.,Dixon, N.E. & Ollis, D.L. (1994) Structure of the Escherichia coli signal transducing protein PII. Structure 2, 981–990. 10. Cudney,R.,Patel,S.,Weisgraber,K.,Newhouse,Y.&McPher- son, A. (1994) Screening and optimization strategies for macro- molecular crystal growth. Acta Crystallogr. D50, 414–423. 11. French, G.S. & Wilson, K.S. (1978) On the treatment of negative intensity observations. Acta Crystallogr. A34, 517. 12. He, L., Soupene, E. & Kustu, S. (1997) NtrC is required for control of Klebsiella pneumoniae NifL activity. J. Bacteriol. 179, 7446–7455. 13. He, L., Soupene, E., Ninfa, A. & Kustu, S. (1998) Physiological role for the GlnK protein of enteric bacteria: relief of NifL inhibition under nitrogen-limiting conditions. J. Bacteriol. 180, 6661–6667. 14. Holm, L. & Sander, C. (1998) Dictionary of recurrent domains in protein structures. Proteins 33, 88–96. 15. Holm, L. & Sander, C. (1998) Touring protein fold space with Dali/FSSP. Nucleic Acids Res. 26, 316–319. 16. Jack, R., de Zamaroczy, M. & Merrick, M. (1999) The signal transducer protein GlnK is required for NifL-dependent nitrogen control of nif gene expression in Klebsiella pneumoniae. J. Bac- teriol. 181, 1156–1162. 17. Jiang, P., Zucker, P., Atkinson, M.R., Kamberov, E.S., Tiraso- phon, W., Chandran, P., Schefke, B. & Ninfa, A.J. (1997) Struc- ture/function analysis of the PII signal transduction protein of Escherichia coli: genetic separation of interactions with protein receptors. J. Bacteriol. 179, 4342–4353. 18. Jones, T.A., Zou, J Y., Cowan, S. & Kjeldgaard, M. (1991) Improving methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47, 110–119. 19. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Throton, J.M. (1993) PROCHECK: a programme to check the sterio- chemical quality of protein structure coordinates. J. Appl. Cryst. A42, 140–149. 20. Little, R., Colombo, V., Leech, A. & Dixon, R. (2002) Direct interaction of the NifL regulatory protein with the GlnK signal transducer enables the Azotobacter vinelandii NifL-NifA reg- ulatory system to respond to conditions replete for nitrogen. J. Biol. Chem. 277, 15472–15481. 21. Magasanik, B. (2000) PII: a remarkable regulatory protein. Trends Microbiol. 8, 447–448. 22. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. (1997) Refine- ment of macromolecular structure by the maximum-likelihood method. Acta Crystallogr. D53, 240–255. 23. Navaza, J.A., (1994) MoRe: an automated package for molecular replacement. Acta Crystallogr. A50, 157–163. 24.Ninfa,A.J.&Atkinson,M.R.(2000)PIIsignaltransduction proteins. Trends Microbiol. 8, 172–179. 25. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac- tion data collected in oscillation mode. Methods Enzymol. 276, 307–326. 26. Pedrosa, F.O., Benelli, E.M., Yates, M.G., Wassem, R., Monte- iro, R.A., Klassen, G., Steffens, M.B.R., Souza, E.M., Chubatsu, L.S. & Rigo, L.U. (2001) Recent developments in the structural organization and regulation of nitrogen fixation genes in Her- baspirillum seropedicae. J. Biotechnol. 91, 189–195. 27. Persuhn, D.C., Steffens, M.B.R., Pedrosa, F.O., Souza, E.M., Yates, M.G. & Rigo, L.U. (2000) The transcriptional activator NtrC controls the expression and activity of glutamine synthetase in Herbaspirillum seropedicae. FEMS Microbiol. Lett. 192, 217– 221. 28. Polikarpov, I., Oliva, G., Castellano, E.E., Garratt, R., Arruda, P., Leite, A. & Craievich, A.F. (1998) The protein crystallography beamline at LNLS, the Brazilian National Synchrotron Light Source. Nucleic Instrum. Methods A. 405, 159–164. 29. Polikarpov, I., Perles, L.A., de Oliveira, R.T., Oliva, G., Castel- lano, E.E., Garratt, R. & Craievich. (1998) Set-up and Experi- mental Parameters of the Protein Crystallography Beamline at the Brazilian National Synchrotron Laboratory. J. Synch. Rad. 5, 72–76. 30. Souza, E.M., Pedrosa, F.O., Drummond, M.H., Rigo, L.U. & Yates, M.G. (1999) Control of Herbaspirillum seropedicae NifA activity by ammonium ions and oxygen. J. Bacteriol. 181, 681– 684. 31. Subramanya, H.S., Roper, D.I., Dauter, Z., Dodson, E.J., Davies, G.J., Wilson, G.J. & Wigley, D.B. (1996) Enzymatic ketonization of 2-hydroxymuconate: specificity and mechanism investigated by the crystal structures of two isomerases. Biochemistry 35, 792–802. 32. Taylor, A.B., Czerwinski, R.M., Johnson, Jr, W.H., Whitman, C.P. & Hackert, M.L. (1998) Crystal structure of 4-oxalocrotonate tautomerase inactivated by 2-oxo-3-pentynoate at 2.4 A ˚ resolu- tion: analysis and implications for the mechanism of inactivation and catalysis. Biochemistry 37, 14692–14700. 33. Thomas, G. Coutts, G. & Merrick, M. (2000) The glnKamtB operon. A conserved gene pair in prokaryotes. Trends Genet. 16, 11–14. 34. Thompson, J.D. Higgins, D.G. & Gibson, T.J. (1994) CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic Acids Res. 22, 4673– 4680. 35. Xu, Y. Carr, P.D. Huber, T. Vasudevan, S.G. & Ollis, D.L. (2001) The structure of the PII-ATP complex. Eur. J. Biochem. 268, 2028–2037. 36. Xu, Y. E.Cheah, P.D. Carr, W.C. van Heeswijk, H.V. Westerhoff, S.G. Vasudevan & D.L.Ollis. (1998) GlnK, a PII-homologue: structure reveals ATP binding site and indicates how the T-loops may be involved in molecular recognition. J. Mol. Biol. 282, 149– 165. Ó FEBS 2002 Structural similarities between PII and GlnK (Eur. J. Biochem. 269) 3303 . The entrance from the top is restricted by Thr31, whose lateral chain is oriented to the interior of the cavity. The interior wall of the cavity is largely hydrophilic Herbaspirillum seropedicae signal transduction protein PII is structurally similar to the enteric GlnK Elaine Machado Benelli 1 ,